1. Field of the Invention
This invention is related generally to techniques for mapping defects on semiconductor surfaces and more particularly to an optical system that can distinguish dislocation pits from grain boundaries in mapping polycrystalline device surfaces.
2. Description of the Prior Art
The quality and suitability of crystalline and polycrystalline materials, such as silicon, gallium-arsenic, and others, for use as substrates for semiconductor applications are affected by defects, such as dislocations, in the crystalline structures. Generally, higher densities of dislocations are indicative of lower quality materials. Therefore, there is a need for systems to detect, measure, and map dislocation densities in crystalline and polycrystalline materials for purposes of analysis and quality control.
There are a number of systems that have been developed to detect and map dislocation densities. The methods used most commonly in the industry currently utilize a surface cleaning or polish step followed by some variation of an etch, which reveals dislocations that intersect the surface of the material by forming a pit at each dislocation site. The pits can then be detected, counted, and mapped, and the density of the pits, i.e., number of pits per unit of surface area, can be determined. This etch pit density (EPD) is considered by persons skilled in this art to be a reliable indicator of the number and density of dislocations in the substrate, and the pits have patterns that reflect slip planes in the crystal lattice of the material.
The most commonly used method of detecting and counting pits to determine EPD is visual observation through a microscope and counting. This process is obviously labor intensive, time consuming, and tedious work. Some alternative systems based on optical technologies to detect and map EPDs have been developed that apparently work on monocrystalline substrates, but none that work reliably for polycrystalline substrates. For example, the U.S. Pat. No. 4,925,298, issued to P. Dobrilla, compares specularly reflected light from an etched sample surface to light reflected from a reference surface to determine EPD. U.S. Pat. No. 5,008,542, issued to D. Look et al. is similar, except it detects light transmitted through the substrate rather than reflected light. In both of those techniques, the specular light is detected, so an increase of dislocation density results in increase of scattered light, and thus a decrease in detection signal. However, polycrystalline substrates present major problems for those systems, because grain boundaries cause substantial light scattering, thus affecting light detection signals and skewing EPD measurements.
It has also been shown and is now well-known in the industry that light scattering from a defect-etched surface can be used to determine surface dislocation numbers statistically. In fact, as reported in B. L. Sopori, "Use of optical scattering to characterize dislocations in semiconductors," 22 APPL OPTICS 4676 (1988), it has been determined that the total integrated light scattered from an illuminated region of a defect-etched surface is proportional to the number of dislocation etch pits in that area, provided that the surface is etched for defect delineation. A light integrating sphere positioned on the surface of the material collects and integrates substantially all of the scattered light, and a photodetector mounted in the integrating sphere measures the integrated light intensity, thus the extent of EPD in an illuminated area on the surface. Again, however, while that large beam statistical EPD detection and mapping technique works well for many applications involving single crystal materials, it does not work well with polycrystalline structures. Grain boundaries in polycrystalline materials are "grooved" in the defect etching process required for defect delineation, so scattering of light by the grain boundaries can cause a larger amplitude integrated light in the integrating sphere, and thus an erroneous EPD signal.